GBE — Generative Being Externalized
Can a small, zero-training detector find structure in streams without knowing what it’s looking at?
This is a runnable test of that question. No neural network. No labels. No domain-specific rules. Just merge, prune, and read the survivors.
Start in 5 Minutes
git clone https://github.com/JackeyLGene/GBE.git
cd GBE
python docs\quickstart_geruon.py
python docs\quickstart_ee_self.py
quickstart_geruon.py— solo Geruon, printstau,F,wit_rate,arrow normquickstart_ee_self.py— 3-cavity Self, prints cross-harm time series- No
pip installrequired. Python 3.8+ stdlib only.
After the quickstarts, read the manuals in order.
Demos
Two runnable demos from the same instrument.
| Demo | Type | Script | Video |
|---|---|---|---|
| Sine Boundary Detection | 3-Cavity Self | demo_self_sine.py | Substack |
| Bach C Major Prelude | Solo Geruon | demo_geruon_bach.py | Substack |
A third bridge experiment — Shepard Paradox — is an honest negative: the instrument is immune to perceptual illusions that lack physical structure.
What This Is
GBE is a centroid detector with memory pressure. Two operations — merge similar frames, prune unreinforced frames — converge streams into structural centroids under finite memory.
| Domain | Core result | Limitations |
|---|---|---|
| UN diplomatic voting | 2025 displacement rank 1/79; falsifiable 2026-2028 | 79 data points; forward prediction unverified |
| Bach WTC | Transposition equivariance 100%; Codex selection loop | Codex operation scaffolded |
| DNA alignments | Exon/intron d=-0.97; AHSG fork d=+1.51 | AHSG: single-gene calibration |
| RNA transcripts | CDS stop 98-100% within three windows | CDS start 66%; passive detection only |
The Trilogy
| Paper | Role |
|---|---|
| I: GEME | Static self-reference |
| II: BGM | Time enters the bridge |
| III: EE | Externalized cognition |
- Paper I: GEME — DOI: 10.5281/zenodo.20344974
- Paper II: BGM — DOI: 10.5281/zenodo.20238099
- Paper III: EE
- EE Supplement
Reproducibility
- Quickstarts: zero dependencies, Python 3.8+ stdlib only.
demo/andexperiments/shepard/scripts are runnable bridge examples.- Frozen experiment scripts under
experiments/. - Domain experiments may require
numpy,pandas,pyarrow, orpybigtools. - Fastest review path: run both quickstarts → skim Geruon Manual → inspect one experiment track.